Granted Patent B2

US 7,693,219 · App. 11/325,066 · Granted Apr 6, 2010

Assignee: Freescale Semiconductor, Inc.

Abstract

Claims (72)

1. A method of encoding video information using fast motion estimation to determine a reference macroblock within a reference frame for a current macroblock in a current frame, comprising:

determining a motion vector, a difference value, and a cost value for each of a plurality of macroblocks within the reference frame according to a fast motion estimation search pattern;

for each of the plurality of macroblocks, comparing cost values and selecting motion vector data, a motion mode and a cost value of a current lowest cost macroblock;

for each of the plurality of macroblocks, determining and storing a difference value for each of a plurality of sub-blocks and at least one corresponding motion vector as stored information;

for each of a plurality of sub-block motion modes, searching the stored information and determining a lowest cost sub-block combination, comparing a cost value of the lowest cost sub-block combination with the cost value of the current lowest cost macroblock;

selecting between the lowest cost sub-block combination and the current lowest cost macroblock as the reference macroblock; and

encoding video information using the selected reference macroblock.

2. The method of claim 1 , further comprising:

comparing relative cost of a plurality of potential start macroblocks;

selecting a lowest cost start macroblock as a first of the plurality of macroblocks within the reference frame; and

determining the fast motion estimation search pattern relative to the selected start macroblock.

3. The method of claim 1 , wherein said determining a difference value for each of a plurality of macroblocks comprises determining a sum of absolute difference (SAD) of picture elements between the current macroblock and a macroblock within the reference frame.

4. The method of claim 1 , wherein said determining a cost value for each of a plurality of macroblocks comprises determining a cost of encoding a motion vector of a macroblock relative to a motion vector prediction and adding the cost of encoding the motion vector to the difference value.

5. The method of claim 1 , wherein said storing a difference value for each of a plurality of sub-blocks and at least one corresponding motion vector as stored information comprises:

storing a motion vector locating a macroblock within the reference frame; and

determining and storing four 8×8 sum of absolute difference (SAD) values for the located macroblock.

6. The method of claim 5 , wherein said searching the stored information and determining a lowest cost sub-block combination for each of a plurality of sub-block motion modes comprises:

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for an upper 16×8 sub-block relative to a first median MV and adding an upper pair of 8×8 SAD values;

determining a lowest cost upper 16×8 sub-block;

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for a lower 16×8 sub-block relative to a second median MV and adding a lower pair of 8×8 SAD values;

determining a lowest cost lower 16×8 sub-block; and

combining a cost of the upper lowest cost 16×8 sub-block with a cost of the lower lowest cost 16×8 sub-block.

7. The method of claim 5 , wherein said searching the stored information and determining a lowest cost sub-block combination for each of a plurality of sub-block motion modes comprises:

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for a left-hand 8×16 sub-block relative to a first median MV and adding a left-hand pair of 8×8 SAD values;

determining a lowest cost left-hand 8×16 sub-block;

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for a right-hand 8×16 sub-block relative to a second median MV and adding a right-hand pair of 8×8 SAD values;

determining a lowest cost right-hand 8×16 sub-block; and

combining a cost of the left-hand lowest cost 8×16 sub-block with a cost of the right-hand lowest cost 8×16 sub-block.

8. The method of claim 5 , wherein said searching the stored information and determining a lowest cost sub-block combination for each of a plurality of sub-block motion modes comprises:

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for an upper-left 8×8 sub-block relative to a first median MV;

determining a lowest cost upper-left 8×8 sub-block;

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for an upper-right 8×8 sub-block relative to a second median MV;

determining a lowest cost upper-right 8×8 sub-block;

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for a lower-left 8×8 sub-block relative to a third median MV;

determining a lowest cost lower-left 8×8 sub-block;

for each of the plurality of macroblocks, determining a cost of encoding a motion vector for a lower-right 8×8 sub-block relative to a fourth median MV;

determining a lowest cost lower-right 8×8 sub-block; and

combining costs of the upper-left, upper-right, lower-left and lower-right 8×8 sub-blocks.

9. The method of claim 1 , further comprising combining a sub-block motion mode bias for each of the plurality of sub-block motion modes.

10. The method of claim 1 , further comprising:

comparing a cost of the current lowest cost macroblock with a predetermined threshold after said selecting motion vector data, a motion mode and a cost value that identifies a current lowest cost macroblock;

comparing a cost of the current lowest cost macroblock with the predetermined threshold after said selecting between the lowest cost sub-block combination and the current lowest cost macroblock to update the current lowest cost macroblock; and

selecting the current lowest cost macroblock as the reference macroblock whenever the cost of the current lowest cost macroblock is less than or equal to the predetermined threshold.

11. A fast motion estimation system that determines a reference macroblock within a reference frame for a current macroblock in a current frame, comprising:

a memory;

a reference macroblock search circuit which determines a motion vector, a plurality of difference values, and a cost value for each of a plurality of macroblocks within the reference frame according to a fast motion estimation search pattern, and which stores said motion vector and said plurality of difference values in said memory;

a sub-block combination search circuit which searches said motion vector and said plurality of difference values in said memory for determining a corresponding one of a plurality of lowest cost sub-block combinations for each of a plurality of sub-block motion modes; and

a comparator circuit which determines a lowest cost macroblock and which selects from among said lowest cost macroblock and said plurality of lowest cost sub-block combinations to determine the reference macroblock.

12. The fast motion estimation system of claim 11 , wherein said comparator circuit compares relative cost of a plurality of potential start macroblocks, selects a lowest cost start macroblock as a first of said plurality of macroblocks within the reference frame, and wherein said reference macroblock search circuit initiates said fast motion estimation search pattern beginning with said lowest cost start macroblock.

13. The fast motion estimation system of claim 11 , wherein said reference macroblock search circuit comprises a sum of absolute difference (SAD) calculation circuit.

14. The fast motion estimation system of claim 11 , wherein said reference macroblock search circuit comprises a motion vector prediction and calculation circuit.

15. The fast motion estimation system of claim 11 , wherein said reference macroblock search circuit determines and stores in said memory four 8×8 sum of absolute difference (SAD) values for each of said plurality of macroblocks within the reference frame.

16. The fast motion estimation system of claim 15 , wherein said sub-block combination search circuit determines a lowest cost upper 16×8 sub-block and a lowest cost lower 16×8 sub-block to determine a lowest cost sub-block combination for a 16×8 motion mode, determines a lowest cost left-hand 8×16 sub-block and a lowest cost right-hand 8×16 sub-block to determine a lowest cost sub-block combination for an 8×16 motion mode, and determines a lowest cost upper-left 8×8 sub-block, a lowest cost upper-right 8×8 sub-block, a lowest cost lower-left 8×8 sub-block, and a lowest cost lower-right 8×8 sub-block to determine a lowest cost sub-block combination for an 8×8 motion mode.

17. The fast motion estimation system of claim of claim 11 , wherein said sub-block combination search circuit combines a corresponding one of a plurality of sub-block motion mode biases for each of said plurality of sub-block motion modes.

18. The fast motion estimation system of claim 11 , wherein said comparator circuit compares a cost value of each of said plurality of macroblocks and each of said plurality of lowest cost sub-block combinations with said predetermined threshold and selects a macroblock or sub-block combination as the reference macroblock if a corresponding cost value achieves said predetermined threshold.

19. A video encoder, comprising:

a first memory storing a current frame;

a second memory storing a reference frame;

a subtract circuit having a negative input receiving interframe prediction information, a positive input receiving current frame information, and an output providing residual information;

a transform and quantization circuit having an input coupled to said output of said subtract circuit and an output providing quantized transform coefficients;

an inverse transform and inverse quantization circuit having an input coupled to said output of said transform and quantization circuit and having an output providing reconstructed residual information;

an adder having a first input receiving said interframe prediction information, a second input coupled to said output of said inverse transform and inverse quantization circuit, and an output providing reconstructed frame information;

a deblock filter having an input coupled to said output of said adder and an output providing filtered reconstructed frame information for storage in said second memory;

a motion compensation circuit having an input receiving at least one reference motion vector of a reference macroblock within said reference frame and an output providing said interframe prediction information; and

a motion estimation circuit having a first input receiving said current frame, a second input receiving said reference frame and an output providing said at least one reference motion vector for said reference block, said motion estimation circuit comprising:

a third memory;

a reference macroblock search circuit which determines a motion vector, a plurality of difference values, and a cost value for each of a plurality of macroblocks within said reference frame according to a fast motion estimation search pattern, and which stores said motion vector and said plurality of difference values in said third memory;

a sub-block combination search circuit which searches said third memory to find a lowest cost sub-block combination for each of a plurality of sub-block motion modes; and

a comparator circuit which determines a lowest cost macroblock and which selects from among said lowest cost macroblock and said plurality of lowest cost sub-block combinations to determine said at least one reference motion vector of said reference macroblock.

20. The video encoder of claim 19 , wherein:

said reference macroblock search circuit stores a 16×16 motion vector and four 8×8 sum of absolute difference (SAD) values in said third memory for each of said plurality of macroblocks within said reference frame according to said fast motion estimation search pattern; and

wherein said sub-block combination search circuit scans said third memory to find a lowest cost sub-block combination for a 16×8 motion mode, a lowest cost sub-block combination for an 8×16 motion mode, and a lowest cost sub-block combination for an 8×8 motion mode.

Assignments (29)

Continuity (1)